Research On Web Service Reliability Evaluation And Improvement

A Web service is characterized by its flexibility to encapsulate discrete business functionalities and its interoperability to support universal application integration. Composition of computational resources and Web-based services into integrated solutions is a key activity to enhance reusability, save cost and design business process from the point of view of customers. Service consumer aims at integrating and reusing existed services, which has greatly changed the process of software development. These changes also influenced the reliability model of Web service compared with traditional software reliability models: traditional software reliability model evaluate it by software faults, but failure factors of Web service also include hardware conditions; traditional software reliability model normally evaluate it during testing phase, but Web service consumer reuse existing service to develop business and expect to evaluate reliability using execution phase feedbacks; traditional software system could not improve its reliability by applying redundant modules, but heterogeneous redundant service could guarantee the enhancement of service reliability.According to those changes, this paper proposes a Web service reliability evaluation and improvement model which has following features,This evaluation model is based on time independent Markov chains. In contrast of other Web service reliability models, it considers reliability of Web services as a measure of failure-free operation from the point of customers'view. It evaluates Web service reliability from failure intensity and recovery intensity to indicate the probability of successful execution.This model is based on client feedbacks. It means that reliability evaluation is in execution phase but not in developing or testing phase. The data collected from this phase could be used to evaluate existing service more effectively and objectively.This model evaluates reliability of Web service system with redundant services. It extends atomic Web service model to node service model, and evaluates reliability for both of them. According to different redundant service types, this paper also suggests a High Reliability First Redundant service selection algorithm to build redundant services for Web service reliability improvement.This paper also designs and implements a prototype system for reliability evaluation model. According to the prototype system, this paper also designs a simulation and experiment scheme to verify our model. Simulation results indicate reliability influenced by different parameters to demonstrate its correctness. Experiment results show reliability improvement by implementing our solutions.